[R] Symbols in R

(Ted Harding) Ted.Harding at manchester.ac.uk
Tue Jun 8 20:27:31 CEST 2010


On 08-Jun-10 18:00:18, Bert Gunter wrote:
> Mount soapbox; begin rant {
> 
> ... However I think it should be added that rarely does this work with
> more than about a half dozen different symbols: a viewer of a graphic
> simply cannot keep the distinctions straight -- or often even decode
> them. Using color to distinguish groups is typically more effective
> (there is actual research to justify this), but of course one then
> runs up against the subjectivity of color perception: aside from the
> obvious (and fairly common) red-green color blindness issue, there's
> also the problem that people perceive different colors differently,
> making some look more prominent than others, which may interfere with
> the decoding of the graphical features.
> 
> In general, in the age of Excel graphics, I would argue that the use of
> point/line characteristics like shape, fill, and color that are decoded
> through legends are much overused and frequently result in visual
> puzzles and illusions, which is the opposite of what a good data
> graphic should do.
> 
> An often better approach is the use of small multiples -- trellis
> graphics -- for which the lattice and ggplot packages provide excellent
> functionality.
> 
> } end rant; dismount soapbox
> 
> Cheers,
> Bert

Hear Hear! Well ranted! About colour, I would add that it is useful
(as Bert hints) only up to the point where the colours remain clearly
distinguishable. When (as one too often does) one sees blobs in
different shades of yellow scattered all over a plot, or in different
shades of purple and magenta, then you cannot see what is what.

You get a similar problem (even with relatively distinct colours)
when the bars of a bar-chart, or the sectors of a pie chart, are in
different colours with the only clue as to what is what being a
"colour key" well off to one side, with no direct visible link to
the graphic, so that the eye has to travel laboriously from one
to the other, searching out the match, and having to consciously
remember which colour is being tracked; by the time you've gone from
chart to key and back again, you have forgotten what it was you looked
at the previous time, so you have been disconnected from the comparison
exercise that is the point of the chart in the first place.

Far better to put the key in a direct visual relationship with the
graphic in the first place (e.g. as labels adjacent to the bars or
sectors, or linked to them by line segments).

Ted
(already booted into rant mode, so "end rant" is an illegal operation)

--------------------------------------------------------------------
E-Mail: (Ted Harding) <Ted.Harding at manchester.ac.uk>
Fax-to-email: +44 (0)870 094 0861
Date: 08-Jun-10                                       Time: 19:27:28
------------------------------ XFMail ------------------------------



More information about the R-help mailing list